An Integration Method of Heterogeneous Models for Process Scheduling Based on Deep Q-Learning Integration Agent

By means of intelligent algorithms and cloud integration paradigm, intelligent manufacturing organizes distributed and heterogeneous resource models organically to realize the process scheduling of complex products. To this end, this paper focuses on the heterogeneous models integration problems in complex production scheduling with machine failure. Specifically, a message bus is established to integrate distributed and heterogeneous cloud resources which include process scheduling rules, meta production process and process execution facilities by uniformly described labeled proxy models. During the whole process scheduling, an virtual intelligent integration agent based on dual deep reinforcement learning is introduced to composite scheduling rules from multiple experts when successively dispatching available process to appropriate machine. Last, the integration method is tested in process scheduling. Compared with integration framework for complete models, this proposed integration paradigm which focuses on interactive models can control system complexity more efficiently. Furthermore, message simulation bus allows for asynchronous communication and log tracking. Compared with the mechanical combination of expert rules, compatible learning algorithm in this paradigm enhances the intelligence of integration.